The “universal suffrage” selection operator, designed primarily for concept learning inside the system REGAL, is discussed for both overlapping and nonoverlapping populations. Analysis of its behavior is performed by using the “virtual average population” method, a new tool for investigating asymptotic properties of convergence of macroscopic quantities related to the population of a genetic algorithm.

Universal suffrage selection, virtual average population, niches and species formation, concept learning
This content is only available as a PDF.
© 1996 by the Massachusetts Institute of Technology
1996
You do not currently have access to this content.